A modular review of the evidence for harmful manipulation by large language models
Current Opinion in Psychology · 2027-04 · 74 · 102478
DOI: 10.1016/j.copsyc.2026.102478 ↗Kurz gefasst
Übersichtsarbeit in der Zeitschrift Current Opinion in Psychology, 2027-04. Verfasst von Jean-François Bonnefon und 1 weiteren Autor:innen. Beteiligte Einrichtung: Toulouse School of Economics, CNRS (TSM-R), Université Toulouse Capitole, Toulouse, France. Electronic address: jean-francois.bonnefon@tse-fr.eu. Die Arbeit nennt 48 Literaturangaben. Gefördert von: Agence nationale de la recherche, UKRI.
Abstract (Originalsprache)
Harmful manipulation by large language models (LLMs) is now a regulatory concern, yet the empirical evidence remains difficult to assess. Harmful manipulation involves multiple features: an influence goal, tactics that bypass reasons or exploit vulnerabilities, limited target awareness, successful influence, and harmful outcomes, which existing studies rarely capture together. This review organizes the literature by progressively stronger constellations of these features. Some studies show that LLMs successfully influence beliefs through overt argumentation, without clear harm. Other studies show that LLMs can influence covertly, but without of harm or bypassing reason. Closest current approximations to harmful manipulation come from studies attempting to operationalize harmful outcomes. Some show little behavioral success, Others capture successful harmful influence, but without covert or reason-bypassing moves. Overall, the literature does not yet provide full evidence of harmful manipulation, but establishes several constituent capacities that make harmful manipulation by LLMs an empirically grounded risk rather than a purely speculative one.
Autor:innen und Institutionen
- Toulouse School of Economics, CNRS (TSM-R), Université Toulouse Capitole, Toulouse, France. Electronic address: jean-francois.bonnefon@tse-fr.eu.
- Inria Bordeaux Sud-Ouest, REGALIA Project-Team, Talence, France; Institute for Ethics in AI, University of Oxford, Oxford, UK.
Förderung
- Agence nationale de la recherche
- UKRI
Zitieren
APA
Bonnefon, J. F., & Prunkl, C. (2027). A modular review of the evidence for harmful manipulation by large language models. Current Opinion in Psychology, 74, 102478. https://doi.org/10.1016/j.copsyc.2026.102478
Vancouver
Bonnefon JF, Prunkl C. A modular review of the evidence for harmful manipulation by large language models. Current Opinion in Psychology. 2027;74:102478. doi:10.1016/j.copsyc.2026.102478
BibTeX
@article{Bonnefon2027,
author = {Bonnefon, Jean-François and Prunkl, Carina},
title = {A modular review of the evidence for harmful manipulation by large language models},
journal = {Current Opinion in Psychology},
year = {2027},
volume = {74},
pages = {102478},
doi = {10.1016/j.copsyc.2026.102478}
}
EUVIMED